Welcome Back, AI Geeks!

It’s time to stay updated with What’s Up in AI.

Sam Altman's spilling the AI tea, Google's flexing with Gemini 2.0 (and the price tag will make you do a double-take), and DeepSeek... well, let's just say their security test results are about as solid as a chocolate teapot. Buckle up - we've got a wild ride ahead! 🎢

In today’s issue:

🤝 Altman’s reveal about AGI costs
🚀 Google waging a literal AI war with Gemini Pro 2.0
🤖 Deepseek fails all security test
⚡ 5 Must-Try AI Tools This Week
📈 What's Trending in AI now
📰 More AI News you need to know

Altman's AI Bombshell: Faster, Cheaper, Bigger

What happened: Sam Altman dropped a detailed blog post breaking down AI's trajectory, highlighting how AI costs are plummeting while capabilities keep climbing. He's painting a picture of an AI future that's both exciting and a bit nerve-wracking, all while subtly teasing a new software engineering agent in the works.

What you should know:

• AI development follows "scaling laws" - throw more resources at it, get better results (though with diminishing returns)

• AI costs are dropping 10x every year, way faster than Moore's Law

• Small improvements in AI intelligence can lead to massive societal benefits

• OpenAI appears to be cooking up a new software engineering agent that could automate code generation and optimization

Why you should care: This isn't just another tech prediction - it's a roadmap from one of AI's biggest players. With AI becoming accessible to everyone and potentially creating massive societal value, we're looking at changes that could reshape everything from work to scientific discovery. Plus, that hint about a new coding agent suggests OpenAI is actively building this future, not just talking about it.

Google launches Gemini Pro 2.0

What happened: After DeepSeek’s R1 and Open AI’s o3 mini, did you seriously think Google would not make its move? A literal war is happening in the AI world of who will become the sorcerer supreme (Marvel reference, if you know you know).
Google just made one of the 'world’s best AI models', Gemini 2.0 (judged on X by LMArena) accessible to everyone. and the numbers are turning heads. Their latest models aren't just dominating the benchmarks - they're crushing the competition on pricing too.

What you need to know:

•  Multimodal capabilities across the entire lineup with a context window to up to 2 million tokens for the Pro Model

• The Power Players:
  - 2.0 Flash: The workhorse model with 1M token context window
  - 2.0 Pro Experimental: Google's best yet for coding and complex tasks
  - 2.0 Flash-Lite: Better than 1.5 Flash but keeps the same speed and cost
  - Flash Thinking: Currently DOMINATING the LLM leaderboards (#1 ranked!)

• At $0.40/1M output tokens, Gemini 2.0 Flash is making waves
  - WAY cheaper than OpenAI's o3-mini ($4.40M)
  - Beats DeepSeek R1's $2.19 pricing
  - Only DeepSeek V3 comes close at $0.28

• Google plans to spend about spend $75 billion this year, much of it on AI
  

  What's Next:

• Image generation and text-to-speech coming soon

• More modalities ready for general availability in months

• Enhanced capabilities for the entire 2.0 family in development

Why you should care: This isn't just another model release - Google's putting serious muscle behind democratizing AI. With features like multimodal input, Google Search integration, and code execution capabilities, they're taking direct aim at both OpenAI and DeepSeek. Plus, with plans to spend $75B this year (mostly on AI), Google's showing they're not backing down from the AI arms race.

With top reasoning models now available to anyone with a Google account, the AI landscape is shifting FAST. The question isn't if OpenAI or Anthropic will respond - it's WHEN.

GIVE GEMINI 2.0 A TRY NOW 
Here’s a prompt to kickstart:

"Hi Gemini 2.0, I need help organizing my day tomorrow to optimize both my productivity and well-being. I'd like you to act as my personal daily planner and life coach. Please ask me any clarifying questions about my priorities, energy levels at different times of the day, and any fixed commitments I already have. Based on my responses, generate a detailed, flexible schedule.
Let’s start by asking me a few questions to tailor the plan perfectly to my needs."

DeepSeek’s AI chatbot just failed every security test thrown at it

In a week that's gone from triumph to trouble, DeepSeek's meteoric rise just hit some serious turbulence. The Chinese AI startup is now facing a perfect storm of security concerns, cyber-attacks, and growing international skepticism. The way DeepSeek has been surprising us every week, y’all better be holding your popcorn for the next time. 

What you need to know:

• DeepSeek actively gathers a broad spectrum of personal and technical data, including:
  1.  Extensive personal info: email, phone, DOB
  2. Stores chat history, text, and audio inputs
  3. Tracks technical data: IP address, keystroke patterns
  4. Retains data for "as long as necessary"
  5. Shares info with service providers and advertising partners

• Countries like Italy, South Korea, Ireland have shown their concern by applying restrictions or removing DeepSeek from app stores. U.S. Navy has warned their personnel against using the tool

• Deep Seek has been recorded to have 100% failure rate in blocking known jailbreaks (tested by Cisco & UPenn)

• Easily can bypass restrictions through simple prompt injections

• Feroot Security(Canadian security company) claims that hidden code of DeepSeek’s can send data to Chinese government

• It allegedly connects to CMPassport.com (China Mobile's registry)

Why this MATTERS?

This isn't just about one app's security issues - it's raising serious questions about AI safety and international data protection. Literally recently a hacker has claimed to have stolen 20 MILLION CREDENTIALS from OpenAI (investigation still going on). With DeepSeek's rapid rise and equally rapid complications, the AI industry is getting a wake-up call about the importance of robust safety measures.

5 AI TOOLS OF THE WEEK

👨‍💻 a0 Dev: Instantly generates full mobile apps with AI, saving developers time and effort.
🗺️ Geospy AI: Analyzes image pixels to determine the exact location a photo was
📉 Trendtracker: Provides AI-driven insights on real-time trending topics for smarter decision-making.
💇 Hair-style.ai: Lets users try different hairstyles on their photos before committing to a real haircut.
🧠 GitHub Copilot: Acts as an AI-powered coding assistant, suggesting code and improving developer productivity.

WHAT’S TRENDING IN AI

• Sam Altman thinks he wont be smarter than gpt-5

• A George Mason professor says OpenAI’s deep research tool is like “having a good PhD-level research assistant, and sending that person away with a task for a week or two” — except it can generate results in as little as five minutes.

•  AI-powered code editor Cursor is now reportedly the fastest app in history to go from $1M to $100M in revenue.

MORE AI NEWS

• Anthropic’s Big AI Plans – At Davos, Anthropic revealed it is developing a two-way voice mode for Claude and plans to launch more advanced models, with over 1 million chips expected to power its AI by 2026.

• Anthropic Challenges Users to Jailbreak Its New AI Model – Anthropic invites users to test the security of its latest AI model by attempting to jailbreak it.

• Kanye West Embraces AI in Music Production – Kanye West confirmed that he's using AI in his upcoming album BULLY, comparing its role in music to autotune.

• Amazon’s $104 Billion AI Investment – Amazon announced plans to spend $104 billion on AI infrastructure in 2025, causing a 3% drop in its stock, though analysts remain optimistic about long-term benefits.

• China Debuts Fully Autonomous Self-Driving Bus – China has introduced a fully autonomous self-driving bus that operates without a steering wheel or driver’s cabin, revolutionizing urban transportation.

• Mistral Expands Le Chat Assistant with New Features – Mistral launched new iOS and Android apps for its le Chat assistant, adding faster responses, enterprise deployments, and expanded functionalities like document processing and code interpretation.

• ByteDance Unveils AI-Powered Deepfake Tool – TikTok owner ByteDance introduced OmniHuman-1, an AI system that creates realistic deepfake videos from just one reference image and audio input.

🚨 Hot Take Alert: Anthropic just banned job applicants from using AI in their resumes and cover letters. Their reasoning? They want to see real writing skills—not AI-generated fluff.

Ironic? Maybe. But here’s the bigger question: If AI literacy is becoming a must-have workplace skill, shouldn’t knowing how to use it effectively matter more than writing a cover letter from scratch?

So, should candidates applying for AI-related jobs be allowed to use AI in their applications?

🚀 Showcase Your Product to 25,000+ AI & Tech Leaders

Put your product in front of CTOs, founders, and senior engineers at companies like Google, Meta, and Amazon. Our highly engaged newsletter audience is actively seeking innovative AI solutions.

Limited spots available for Q1 2025 → Get in touch with us here.

Help us get better and suggest new ideas that you’d like to know more about at [email protected]

Give us a quick pulse check for today’s newsletter 👇️ 

We’ll be back tomorrow with a deep dive on making AI-Powered PPTs in just few seconds using a mind-blowing app! Subscribe here so you don’t miss out.

Stay Tuned! 🤖 
Team What’s Up in AI